Load and prepare data

Load data as xts object:

ts<-readRDS('pumpTemp.rds')

Reduce data

require("xts")

## Loading required package: xts

## Loading required package: zoo

## 
## Attaching package: 'zoo'

## The following objects are masked from 'package:base':
## 
##     as.Date, as.Date.numeric

time(ts) <- round(time(ts), "mins")
ts<-xts::first(ts, n=1440*60)
ts<-xts::last(ts, n=1440*14)

Savitzky Golay

require("prospectr")#savitzky golay

## Loading required package: prospectr

## Loading required package: RcppArmadillo

distanceCrit<-1
ts2<-ts
width <- 1439
smoothTs <- t(savitzkyGolay(t(as.data.frame(ts2)), p=9, w=width, m=0))
smoothTs<-xts(smoothTs,as.POSIXct(rownames(smoothTs), tz = "GMT", origin = "1970-1-1"))
names(smoothTs) <- 'smoothTs'
smoothTs0 <- t(savitzkyGolay(t(as.data.frame(ts2)), p=0, w=143, m=0))
smoothTs0<-xts(smoothTs0,as.POSIXct(rownames(smoothTs0), tz = "GMT", origin = "1970-1-1"))
names(smoothTs0) <- 'smoothTs0'

align time (savitzkyGolay cuts the first elements) by adding dummy data:

tmpIndex <- 1:((width-1)/2)
dummyData <- ts2[tmpIndex, ]
dummyData[ ,1] <- NA
smoothTs <- rbind(smoothTs, dummyData)

require("dygraphs")

## Loading required package: dygraphs

dygraphs::dygraph(merge(ts2, smoothTs, smoothTs0))

If noise is really to be flagged (rahter than using the de-noised signal created by a de-noising routine as shown above) then sav gol still seems more useful than moving average:

my.outlier = ts2
flag <- which(abs(ts2 - smoothTs) > distanceCrit)
mode(my.outlier) <- "logical";
my.outlier[flag] <- FALSE
my.outlier[-flag] <- TRUE

my.outlier0 = ts2
flag0 <- which(abs(ts2 - smoothTs0) > distanceCrit)
mode(my.outlier0) <- "logical";
my.outlier0[flag0] <- FALSE
my.outlier0[-flag0] <- TRUE

require("m181")

## Loading required package: m181

m181::visualise_dy(ts2, my.outlier)

m181::visualise_dy(ts2, my.outlier0)